The conclusion we arrived at in the previous section is that Force is a phase head, and complements that lack Force are not phasal. Concomi-tantly, those non-phasal complements lack Top and Foc as well on the assumption by Haegeman (2006a) that Force licenses Top and Foc. As opposed to this truncation analysis, one may wonder if the absence of the-matic topics and obligatory exhaustive foci in complement clauses entails the impoverishment of their projections, given the operator movement analysis proposed by Haegeman (2007; 2009; 2010a;b) and Haegeman–
Ürögdi (2010) (henceforce HÜ). This section aims to provide further evidence in support of the truncation analysis and against the operator movement analysis.
In a series of recent works, Haegeman (2007; 2009; 2010a;b) modifies her own truncation analysis, and suggests that all complement clauses have a full structure like (59a). Instead of clause reduction, she argues that there is movement of an operator to the left periphery as in (61)
23A reviewer questions how tense is analyzed in the present analysis if a tense feature is inherited from C. I follow Chomsky (2007, 20) in supposing that tense is an inherent property of T, and an interpretable tense feature originates on T. Tense is not directly affected by the internal structure of CP. For example, subjunctive and factive complements have the reduced structure in (59b), but they are tensed.
that gives rise to an intervention effect with argument fronting such as topicalization and focalization.
(61) [CP Opi C [FPti [TP . . . ]]] (Haegeman 2007)
Empirical support for operator movement comes from wh-movement in temporal clauses. In this light, consider (62).
(a)
(62) *When this column she started to write last year, I thought she would be fine.
(b) *When this song I heard, I remembered my first love.
(c) When last year she started to write this column, I thought she would be fine.
(Haegeman 2010a, 597)
(62) shows that there is an asymmetry between argument fronting (62a,b) and adjunct fronting (62c). Based on analyses which assume that tempo-ral adverbial clauses are derived bywh-movement of a temporal operator to the left periphery (Geis 1970; Larson 1987; 1990; Johnson 1988; Demir-dache–Uribe-Etxebarria 2004 among others), she concludes that the op-erator in the left periphery intervenes with argument fronting but it does not do so with adjunct fronting. As shown in (63), this asymmetry be-tween argument fronting and adjunct fronting is independently detected in relative clauses that involve operator movement.
(a)
(63) These are the students who in the next semester will study these texts.
(b) *These are the students who these texts will study in the next semester.
(ibid.)
The point of concern is that the same asymmetry is observed in factive complements:
(a)
(64) I regret that in those days I didn’t realize the importance of classical languages.
(b) *John regrets that this book Mary read. (Maki et al. 1999, (2c))
Based on this observation and others, HÜ conclude that there is also operator movement in factive complements that gives rise to an interven-tion effect, and that this operator makes factive complements referential in the same way as DPs (see also de Cuba–Ürögdi 2009).
In Japanese, no such asymmetry can be found in environments where operator movement is supposed to be involved. As shown in (65), no
thematic topicalization of either an argument or an adjunct is allowed in the relative clause.
(a)
(65) *Ken-wa [NP [TP sono hon-wa kinoo yon-da] hito]-o home-ta.
Ken-top that book-top yesterday read-past person-acc praise-past
‘Ken praised the person who as for that book, s/he had read it yesterday.’
(b) *Ken-wa [NP [TP kinoo-wa sono hon-o yon-da] hito]-o home-ta.
Ken-top yesterday-top that book-acc read-past person praise-past
‘Ken praised the person who as for yesterday, s/he had read that book.’
How, then, do we know whether there is operator movement in Japanese?
Note that both the truncation analysis and the operator movement anal-ysis would predict that argument fronting (e.g., thematic topics and obligatory exhaustive foci) is disallowed. So, argument fronting alone can-not be decisive evidence to determine which analysis is more appropriate.
Rather, what is crucial is the existence of a weak island as shown in (66).
(66) *Why did you notice that Maria fixed the cart? (HÜ, (15b))
The operator movement analysis can account for the weak island in (66) because wh-movement across the operator in the left periphery of the embedded clause contravenes Relativized Minimality (Rizzi 1990; 2004).
On the other hand, the truncation analysis I employ here cannot account for (66) because there is nothing to prevent movement of the wh-phrase to the front of the sentence. So, by examining whether a relevant clause constitutes a weak island, we are able to know which analysis is suitable for Japanese.
In fact, Hiraiwa (2010b) argues that factive complements in Japanese are weak islands, and illustrates this claim in terms of the following data:
(a)
(67) Ken-ga [Naomi-ga dare-kara-mo okane-o moraw-anakat-ta Ken-nom Naomi-nom who-from-even money-acc receive-neg-past koto-o/no-o] sira-nakat-ta sooda.
C-acc/C-acc know-neg-past hear.say
‘I heard that Ken didn’t know that Naomi didn’t receive money from anyone.’
(b)??Dare-karai-mo Ken-ga [Naomi-ga tiokane-o moraw-anaka-ta who-from-even Ken-nom Naomi-nom money-acc receive-neg-past koto-o/no-o] sira-nakat-ta sooda.
C-acc/C-acc know-neg-past hear.say
‘I heard that Ken didn’t know that Naomi didn’t receive money from anyone.’
(Hiraiwa 2010, (12))
Furthermore, following Watanabe (1996), Hiraiwa maintains that the ad-nominal form of predicates arises from operator movement, since the adnominal form typically appears in relative clauses that involve opera-tor movement. Recall that in section 3we saw that complements headed bykoto/no require a predicate in an adnominal form, and that (true and semi-)factive clauses are one such complement. The relevant example is repeated in (68).
(68) Keni-wa [Naomi-ga zibuni-o suki-na/*da koto-o/no-o] wasure-ta/yorokon-da.
Ken-top Naomi-nom self-acc like-pres/pres C-acc/C-acc forget-past/be.glad-past
‘Keniforgot/was glad that Naomi liked himi.’
Thus, if the adnominal form is a manifestation of operator movement, it follows that factive complements constitute a weak island because of operator movement. However, things do not work the same way for sub-junctive complements. As noted in section 2, subjunctive complements can also be headed by koto, and so a predicate ends in an adnominal form. The relevant example is (69).
(69) Keni-wa [Naomi-ga zibuni-o suki-na/*da koto]-o negat-ta.
Ken-top Naomi-nom self-acc like-pres/pres C-acc wish-past
‘Keniwished that Naomi would like himi.’
If the adnominal form results from operator movement, it is predicted that subjunctive complements will also constitute a weak island. However, this prediction is not borne out. As shown in (70), extraction of a wh-phrase out of a subjunctive complement is not as bad as (67b).
(a)
(70) Ken-ga [Naomi-ga dare-kare-mo okane-o moraw-ana-i Ken-top Naomi-nom who-from-even money-acc receive-neg-pres koto]-o negat-ta sooda.
C-acc wish-past hearsay
‘I heard that Ken wished that Naomi would not receive money from anyone.’
(b)?Dare-karai-mo Ken-ga [Naomi-ga tiokane-o moraw-ana-i who-from-even Ken-nom Naomi-nom money-acc receive-neg-pres koto]-o negat-ta sooda.
C-acc wish-past hearsay
‘I heard that Ken wished that Naomi would not receive money from anyone.’
Accordingly, this suggests that the adnominal form of predicates does not necessarily entail operator movement. Nonetheless, one could still argue
that operator movement is involved only in factive complements as they constitute a weak island. However, in the face of the literature arguing that factive islands are a semantic/pragmatic rather than a syntactic phenomenon (Szabolcsi–Zwarts 1993; Oshima 2007 among many others), the motivation for operator movement is weakened.24The bigger problem is that it is unclear why thematic topics and obligatory exhaustive foci are precluded in non-phasal complements other than factives. As just noted, subjunctive complements do not constitute a weak island; and neither do control complements as shown in (71) and (72).25
(a)
(71) Keni-ga [PROi/*jdare-kara-mo okane-o moraw-ana-i Ken-nom PRO who-from-even money-acc receive-neg-pres yoo(ni)/koto-o] keikakusi-ta sooda.
C/C-acc plan-past hearsay
‘I heard that Ken planned not to receive money from anyone.’
(b) Dare-kara-mo Keni-ga [PROi/*jtokane-o moraw-ana-i who-from-even Ken-nom PRO money-acc receive-neg-pres yoo(ni)/koto-o] keikakusi-ta sooda.
C/C-acc plan-past hearsay
‘I heard that Ken planned not to receive money from anyone.’
(a)
(72) Ken-ga Naomii-ni [PROi/*j dare-kara-mo okane-o Ken-nom Naomi-dat PRO who-from-even money-acc moraw-ana-i yoo(ni)/koto-o] susume-ta sooda.
receive-neg-pres C/C-acc recommend-past hearsay
‘Ken recommended Mary not to receive money from anyone.’
(b)?Dare-kara-mo Ken-ga Naomii-ni [PROi/*j tokane-o who-from-even Ken-nom Naomi-dat PRO money-acc moraw-ana-i yoo(ni)/koto-o] susume-ta sooda.
receive-neg-pres C/C-acc recommend-past hearsay
‘Ken recommended Mary not to receive money from anyone.’
If weak islands are a hallmark of operator movement, this suggests that subjunctive and control complements do not involve operator movement.
Thus, the operator movement analysis is not able to account for the
24See also Basse (2008) for an alternative to the operator movement analysis of factive islands.
25It appears that copulas and nominal adjectives that inflect for an adnominal ending cannot readily occur in control complements.
absence of thematic topics and obligatory exhaustive foci in these com-plements.26 Viewed in this light, it seems reasonable to conclude that thematic topics and obligatory exhaustive foci are disallowed in these complements for reasons other than operator movement.
In contrast, the truncation analysis can account for this by assuming that these complements lack Top and Foc projections. Furthermore, the advantage of the truncation analysis is that it can uniformly account for the link between phasehood and Top/Foc, whilst an operator has nothing to do with phasehood. Accordingly, in the absence of decisive evidence for operator movement in Japanese, I assume that the absence of thematic topics and obligatory exhaustive foci indicates the lack of Top and Foc projections.